Interview
Julian Hazell – Off the Clock with 80k #4
- The UK has officially entered a recession, a development the speakers note with a mix of dark humor and criticism of government economic management.
- Julian Hazel, a Canadian content writer and guest, jokes about his impending move to North America and the potential "betrayal" of his British identity regarding US/UK cultural differences.
- The group discusses the British royal family's transition to King Charles, with Hazel noting the lack of "progessive" monarchy compared to the recent Queen Elizabeth II era.
- A tangent on Calvinism and religious determinism leads to a discussion on "Evidential Decision Theory," where Julian contrasts the view that good acts signal pre-saved status against Causal Decision Theory.
- Julian recounts a detailed conversation with a Calvinist missionary whose life goal is to "save souls," despite the theological implication that the saved were always destined for salvation.
- The speakers debate whether such Calvinists implicitly endorse compatibilism to make sense of their own agency, contrasting this with Sartre's concept of "bad faith."
- The conversation shifts to the "One-Boxer vs. Two-Boxer" thought experiment in decision theory, where a perfect predictor places $1 million in an opaque box only if it predicts the agent will take only that box.
- Julian chooses to take only the opaque box (one-boxing), citing a resignation to the prediction mechanism, while others note the counter-intuitive nature of forgoing guaranteed extra money.
- Variations of the thought experiment are explored, including a "grandpa in the attic" scenario and a "desert rescue" scenario involving a promise to pay a rescuer, highlighting tensions between rational choice and virtue ethics.
- Julian describes his "origin story" for effective altruism, sparked by a Reddit post questioning the utility of classical philosophy (e.g., reading Aristotle) versus modern science.
- The group defends Julian's "oafish" questioning against humanities academics, arguing that practical philosophy (e.g., Peter Singer, Will MacAskill) has tangible impact, unlike academic exegesis.
- The speakers criticize the peer-review process in academia, citing a viral case where a journal accepted a paper containing a DALL-E-generated image of oversized mouse testicles with gibberish text.
- They note the double standard where academics fiercely defend peer review despite its frequent failures, while often dismissing pre-prints (like those on arXiv) in fields like AI/ML.
- Concerns are raised regarding the reproducibility crisis in LLM research due to proprietary models and the inability to replicate experiments on updated or fine-tuned systems.
- Cody Fenwick discusses a legal paper claiming LLM outputs are not protected speech under the First Amendment, arguing the claim is an overreach that ignores core political speech protections.
- A critique of the US law review system reveals that second-year law students grade and edit papers by professors, a practice the speakers find absurd compared to the UK's anonymous, centralized exam system.
- The group compares UK and US/Canadian higher education systems, noting the UK's rigid degree structures limit "shopping" for courses, whereas the US allows flexibility that can inflate GPAs at less selective institutions.
- Cody Fenwick shares his anecdote of attending a low-prestige undergraduate institution to achieve a high GPA, which facilitated his acceptance into Harvard Law School despite a modest LSAT score.
- The speakers discuss the "gameability" of the US admissions system, where students may exploit the incentive structure of easier grading at less prestigious schools to boost their credentials.
- The group confirms an "ADK house view" of opposing the assassination of King Charles, treating the topic with humorous gravity regarding the "act/omission" distinction.
- The conversation concludes with a brief debate on the public discourse surrounding AI safety, specifically the amplification of extreme views by figures like Nick Bostrom and Eliezer Yudkowsky versus more moderate expert consensus.
- Julian defends the utility of Harry Potter and the Methods of Rationality as a compelling entry point to rationalist thought, despite its fictional nature.
- The speakers note that public perception of AI risk is currently shaped by "angriest" voices on social media, potentially distorting the median view of experts who acknowledge risk but express uncertainty about timelines.
- The group identifies "historical fencing" (HEMA) as a serious sport, joking that practitioners should be granted automatic voting rights due to their demonstrated "democratic fitness."